Real-World Clinical Outcomes of Pembrolizumab as Second-Line Therapy for Advanced Urothelial Carcinoma

Background: Pembrolizumab is currently considered the standard second-line treatment for advanced urothelial carcinoma (UC). This study aimed to investigate the ecacy and safety of pembrolizumab in patients with advanced UC in real-world data, which is not well-reported. Methods: The study included 98 patients with advanced UC who were treated with pembrolizumab after platinum-based chemotherapy at eight hospitals. The ecacy, safety, and risk factors for prognosis were evaluated. Results: The median age was 73 years. Nineteen patients (19%) with performance status (PS) 2-4 were included. The percentages of liver, lung, bone, and lymph node metastasis were 18%, 27%, 19%, and 76%, respectively. The best response was complete response in nine patients (9%) and partial response in 16 patients (17%). The median progression-free survival and overall survival were 3.7 months (95% condence interval [CI]: 2.8–4.7) and 11.8 months (95% CI: 6.7–17.0), respectively. Poor PS, liver metastasis, and higher C-reactive protein were poor prognostic factors. Hyperprogression was observed in nine patients (9%), who were mostly of poor PS and had high-volume lesions. Severe adverse events (AEs) were observed in 18 patients (19%), and ve patients died because of AEs (5%). Atypical AEs, such as pneumocystis pneumonia and acute myeloid leukemia, were observed in our cohort. Conclusion: The ecacy of pembrolizumab for advanced UC in our cohort was better than a previous phase III trial (KEYNOTE-045), possibly owing to lower cancer volume and fewer visceral metastases in our patients.

Background M-VAC (methotrexate, vinblastine, doxorubicin and cisplatin) chemotherapy [1], and gemcitabine and cisplatin chemotherapy for advanced urothelial carcinoma (UC) has been used for a long time [2].
Platinum-based chemotherapies as rst-line treatment for advanced UC are effective, but few cases have achieved complete remission. For a long time, there had been no effective second-line treatment after failure of rst-line chemotherapy for advanced UC. In a randomized, phase 3 study (KEYNOTE-045) for patients with advanced UC after failure of platinum-based chemotherapy, pembrolizumab resulted in longer overall survival (OS) than existing second-line chemotherapy (10.3 months vs 7.4 months) and achieved a 27% risk reduction of death [3]. Pembrolizumab is a highly selective, humanoid monoclonal antibody against programmed death 1 (PD-1) [3]. In addition, the incidence of treatment-related adverse events (AEs) was lower with pembrolizumab, including fewer high-grade events (grade 3-5) AE that resulted in discontinuation of the treatment [3]. However, in the cited clinical trial, inclusion of participants was limited by cancer and patient status, such as measurable lesions prescribed by the Response Evaluation Criteria in Solid Tumors (RECIST) and the Eastern Cooperative Oncology Group performance status (ECOG-PS) [3]. Therefore, we need data on pembrolizumab for advanced UC in real-world clinical practice. In this study, we retrospectively investigated the e cacy and safety of pembrolizumab in patients with advanced UC in real-world data from Kanazawa University and related hospitals.

Patients
We retrospectively evaluated advanced and recurrent UC patients who were treated with pembrolizumab after progression of platinum-based chemotherapy at our and related hospitals from January 2018 to October 2019.

Treatment and follow-up examination
Pembrolizumab was administered at a dose of 200 mg every three weeks until discontinuation due to disease progression or unacceptable AEs. We obtained medical records, and clinical features were recorded, including age, sex, smoking status, primary tumor site, histology, metastatic organs, duration since previous chemotherapy , and ECOG-PS. Laboratory data including hemoglobin (Hb) level, C-reactive protein (CRP) level, neutrophil counts, and lymphocyte counts were also collected from medical records.
Neutrophil-lymphocyte ratio (NLR) was the ratio of neutrophils to lymphocytes. The Bellmunt risk score was de ned as the number of clinical factors composed of worse ECOG-PS (>1), presence of liver metastasis, short duration since previous chemotherapy (<3 months), and low Hb (<10 mg/dL) [3].
Treatment e cacy was assessed using RECIST version 1.1 [4]; however, for non-measurable lesions evaluated by RECIST criteria, such as small lesions or bone metastasis, the total diameter of a lesion with 20% (and over 5mm) increase in size or a new lesion were de ned as progressive disease (PD); the total diameter of the lesions with a 30% decrease in size were de ned as partial response (PR); and lesions with 30% decrease in size and coinciding with RECIST criteria were de ned as complete response (CR), which indicated disappearance of all measurable lesions. The objective response rate (ORR) was de ned as the proportion of PR and CR, and disease control rate (DCR) was de ned as the proportion of PR, CR, and stable disease (SD). Progression-free survival (PFS) was calculated from the commencement of pembrolizumab to radiographic or clinical disease progression or death. Durable response rate (DRR) and duration of response (DOR) from the commencement of pembrolizumab to radiographic or clinical disease progression or death of CR and PR patients were also calculated. OS was calculated from the commencement of pembrolizumab to death or the time of last follow-up. Pseudoprogression (PP) was de ned as a response to treatment after an initial increase in volume of cancer lesions or with the appearance of new lesions subsequently followed by disease stabilization or a disease response before con rming the progression with a second assessment using imaging evaluation [5]. Hyperprogression (HP) was de ned as rapid tumor progression after the initiation of pembrolizumab [6]. Toxicity, including immune-related adverse events (irAE), was assessed according to the Common Terminology Criteria for Adverse Events (CTCAE) version 4.0 [7].

Statistical analysis
Categorical variables were used to calculate the incidence and percentage of each factor, and all continuous variables were reported as median and range. PFS and OS were estimated using Kaplan-Meier methods, and the differences were evaluated using the log-rank test. Cox proportional hazards models were used to identify the prognostic factors for OS. Correlations within each variable were calculated using Spearman's rank correlation, and the correlation coe cients were then interpreted following the guidelines proposed by Cohen, a small correlation being 0.1-0.3, medium 0.3-0.5, and large 0.5-1.0 [8]. All data analyses were performed using SPSS for Windows (SPSS Inc., Chicago, IL), taking p-value<0.05 as statistically signi cant.

Results
The study included 98 patients with advanced UC who were treated with pembrolizumab after platinumbased chemotherapy at Kanazawa University and seven related hospitals. The characteristics of the patients are presented in Table 1. Fifteen (16%) patients had no measurable lesion evaluated by RECIST (a small lesion, non-measurable bone metastasis, or non-measurable original tumor). Nineteen (20%) patients were ECOG-PS 2-4.
We investigated variables that predict shorter OS. In univariate analysis, poor PS (2-4), liver metastasis, lung metastasis, two or more metastatic organs, low Hb levels, higher CRP, and higher NLR were identi ed as factors associated with short OS (Table 2). CRP had medium negative correlation with Hb (R = − 0.458, p < 0.001) and strong positive correlation with NLR (R = 0.622, p < 0.001), respectively. With regard to metastatic site, liver metastasis had the largest hazard ratio in the univariate analysis. Therefore, we performed multivariate analysis using age, liver metastasis, and CRP. As a result, each parameter was statistically signi cant (Table 2). Poor PS and liver metastasis were strong predictors for poor prognosis in our study. The OS of patients with PS 0-1 and no liver metastasis was signi cantly longer than each PS 0-1 and liver metastasis (p < 0.001), PS 2-4 and no liver metastasis (p < 0.001), and PS 2-4 and liver metastasis (p < 0.001), respectively (Fig. 2).
The pembrolizumab treatment-related AEs are shown in Table 3. Severe AEs (including grade 3-5 AEs) were observed in 18 (19%) patients. Interstitial pneumonia was the most frequent grade 3-5 AE in this study. Other types of pneumonia included one pneumocystis pneumonia and one bronchial pneumonia. Five (5%) patients died after pembrolizumab treatment because of AEs. Two (2%) patients died of interstitial pneumonia, one (1%) of pneumocystis pneumonia, one (1%) of ulcerative colitis, and one (1%) of AML, respectively. Two (2%) patients with persistent bone marrow suppression caused by previous chemotherapy suffered with AML after two and four administrations of pembrolizumab, respectively.

Discussion
We report the clinical outcome after pembrolizumab in Japanese patients with advanced UC. The median PFS and OS were 3.7 months and 11.8 months, respectively, and ORR was 26%. These outcomes were slightly better than the KEYNOTE-045 trial [3]. This might be because of the patients' backgrounds. Poor prognostic factors after systemic treatment for advanced UC are poor PS and visceral metastases [9][10][11][12], especially liver metastases [13]. In the KEYNOTE-045 trial, the proportion of PS 0-1, PS 2, and PS 3-4 participants were 97.1%, 2.9%, and 0%, respectively [3]. In our study, the proportion of patients with PS 2-4 was 20 . The evaluation of risk factors for OS revealed poor PS and badly-affected OS. The median OS of patients with PS 0-1 and PS 2-4 was 17.7 months and 1.4 months, respectively (data not shown). The inclusion criteria for KEYNOTE-045 trial for cancer status determined at least one measurable lesion evaluated by RECIST version 1.1 [3]. The measurable lesions evaluated by RECIST was a tumor with a minimum of 10 mm long diameter or a lymph node with a minimum of 15 mm short diameter [4]. In our study, 15 cases (16%) did not have the lesions evaluated by the RECIST. The visceral metastases and liver metastases, which were reported as poor prognostic factors [9][10][11][12][13], were 62% and 18% in our study, compared to 89% and 34% in KEYNOTE-045 [3]. In our study, patients with good PS 0-1 and fewer visceral metastases, which were good prognostic factors for OS (Table 2), would contribute to good outcomes of our study. The OS of the patients with PS 0-1 and no liver metastasis was signi cantly better than that of the patients with PS 2-4 or liver metastases (Fig. 2).
We also evaluated the e cacy of pretreatment blood tests on prognosis. In ammatory mediators are produced to create a cancer-related in ammatory microenvironment that increases cancer aggressiveness related to invasion, enhanced angiogenesis, immune system resistance, and resistance to therapy [14]. In ammatory markers, such as CRP, used as surrogates of systemic in ammation, are reported to be poor prognostic factors for advanced UC [15]. NLR is an index of systemic in ammation and a biomarker of aggressiveness. A high NLR before treatment has been reported to indicate a heightened neutrophil-dependent in ammatory reaction and decreased lymphocyte-mediated antitumor immune responses [16]. An increased pretreatment NLR in patients with UC has been reported to predict poor OS, PFS, and cancer-speci c survival in a systematic review and meta-analysis [17]. In our study, we investigated Hb, CRP, and NLR. Our results indicate that high preoperative in ammatory markers were associated with poor prognosis after pembrolizumab treatment.
After treatment with immune checkpoint inhibitors, atypical patterns of response, such as PP and HP, were reported. The PP is characterized by a transient increase followed by a decrease in tumor size [5]. The histopathological ndings of the lesion biopsies revealed the presence of in ammatory in ltration or necrosis; therefore, PP is possibly associated with the in ltration of active T cells and other immune cells at the lesion [18]. Rosenberg et al reported that PP was observed in 20 (6%) out of 310 patients with advanced UC who were treated with Atezolizumab after progression of platinum-based chemotherapy [19]. Sharma et al reported that PP was observed in 24 (9%) out of 265 patients with advanced UC who were treated with Nivolumab after progression of platinum-based chemotherapy [20]. In our study, we observed PP in three (3%) patients. HP is de ned as rapid increase in tumor growth rate [21]. There were few reports HP of UC. Hwang et al reported HP in patients with UC treated with PD-1/PD-L1 inhibitors were observed in 12 (11.9%) out of 101 patients [22]. Higher CRP, higher neutrophil count, and higher volume of target lesions were reported to be associated with increased risks of HP [23]. In our study, 9 patients after pembrolizumab treatment progressed rapidly, and these patients de ned as HP. All patients had multiple lesions with poor ECOG PS. Eight patients were PS2-4, and the other, who was PS 1, had liver, brain, peritoneal, and lymph node metastases without removal of primary lesion. Median OS of these 9 patients was 29 days (range 11-43 days). All patients died before rst imaging evaluation schedule; therefore, there were no imaging data to evaluate actual progressions of the cancer, and it was di cult to distinguish these cases as HP or natural course. In our study, the prognosis of patients with PS 2-4 was poor (Fig. 2). In addition, patients with PS 2-4 and high-volume lesion would be expected not only poor response of pembrolizumab treatment but also probable HP; therefore, we should be careful to administer pembrolizumab to the patients.
In our study, 51 (53%) patients experienced AEs, including 18 (19%) patients with severe AEs (grade 3 or more) and 5 (5%) with death (grade 5). The frequency of AEs in our study was higher than that in the KEYNOTE-045 trial [3]. The median age of our study, which was 73 years old, was higher than the median age, which was 67 years old [3], might affect the proportion of AEs [24]. Interstitial pneumonia was observed in 10 patients (10%) in all grade and 8 patients (8%) in grade 3-5. In our study, 5 patients died after pembrolizumab treatment. One of them was Pneumocystis Jirovecii pneumonia. The cases with Pneumocystis Jirovecii Pneumonia caused by immunosuppression for Immune checkpoint-related toxicity were reported [25]. Therapy-related myelodysplastic syndrome or AML with chemotherapy for solid cancer was well known [26]; however, a report of AML after immune checkpoint inhibitor was not found. The two cases of AML after pembrolizumab treatment received each 26 and 27 courses of platinum-based chemotherapy before pembrolizumab commencement and suffered chronic bone marrow suppression at the commencement of pembrolizumab; therefore, it was unclear that the onsets of AML were natural course after long term chemotherapy or irAE.
Limitations of our study are retrospective design, small cohort, and short follow-up period. Our study revealed the feature of pembrolizumab treatment for advanced UC current clinical practice in Japan. We could discuss the difference between clinical trial and real clinical practice; therefore, we also investigated the cohort with inadaptable background to clinical trial, such as patient with poor PS and low volume or unmeasurable cancer status evaluated by RECIST, in this study. Conclusions E cacy of pembrolizumab for advanced UC of our cohort was better than previous phase III trial, KEYNOTE-045, probably because of lower cancer volume and fewer visceral metastasis of our patients. We demonstrated that poor PS, liver metastasis, and high pretreatment CRP were poor prognostic factors in patients with advanced UC treated by pembrolizumab. HP was observed mostly in poor PS patients with high volume lesions. Untypical AEs, such as, Pneumocystis Pneumonia and AML, were observed of our cohort. This study was small size pilot study; therefore, large size and longtime follow-up data to clarify our outcomes should be conducted.

Availability of data and materials
Because of ethical restrictions, the raw data underlying this paper are available upon request to the corresponding author.

Ethics approval and consent to participate
This study was approved by the ethics review board of Graduate School of Medical Science, Kanazawa University (No. 2018-082, 2847). Informed consent was obtained in the form of opt-out on the web-site.

Consent for publication
We obtained the ethics review board approval, including publication allowance.